Flat or Percentage Staking: How Your Bankroll Changes

Flat staking keeps the cash stake fixed until a reset defined in advance. Percentage staking keeps the stake at a fixed share of your current bankroll, so the cash amount changes after each settled bet. Neither method makes negative expected value (EV) bets profitable or outperforms the other under every condition.

Your bankroll is money set aside for betting that you can afford to lose. It excludes borrowed money and funds needed for essential expenses. This guide compares cash stakes on sports outcomes only.

What stays fixed under each method?

Flat staking fixes a cash amount; percentage staking fixes a share of a defined bankroll base. Identical starting stakes therefore do not mean identical risk as the balance changes.

With a $1000 bankroll, both methods can start at $20: a fixed $20 stake or 2% of the bankroll. After one win at decimal odds of 2, the next percentage stake is $20.40. After one loss, it is $19.60. The fixed stake remains $20.

Here and throughout this guide, 2% is an example parameter—not a recommendation or a proven safe or optimal stake size.

How do the methods compare on the same outcome sequence?

On the same ten-bet sequence, flat staking finishes at $1040 and percentage staking at approximately $1038.74. This is one specified example, not an estimate of a typical result.

The starting bankroll is $1000, the fixed stake is $20, the percentage stake is 2% of the current bankroll, and decimal odds are 2. Each bet settles fully before the next one. The sequence is win, win, loss, win, loss, loss, win, win, loss, win.

This calculation excludes fees, deposits, withdrawals, correlated outcomes, overlapping bets, minimum stakes, and stake-rounding constraints. Internal values are not rounded after each step; the table displays amounts to the nearest cent.

The unrounded percentage-staking ending balance is 1038.7363585176843. The next stake is 20.774727, or about $20.77. Net profit is $40 with fixed stakes and approximately $38.74 with percentage stakes—not the entire ending bankroll.

The chart shows only the specified outcome sequence. It is not a median, a random simulation result, or historical bookmaker data. The curves are very close in this example, especially on mobile; read the exact small differences from the preceding table or the chart tooltip.

To check the calculation, let B0 be the starting bankroll, d the decimal odds, s0 the fixed stake, f the fraction of the current bankroll, W the number of wins, and L the number of losses. A win adds net profit of s(d−1); a loss subtracts the stake s.

• For fixed stakes: B = B0 + s0[W(d−1)−L], while the bankroll can cover the stake. Otherwise, betting must stop or the stake must be limited.
• For percentage stakes: B = B0[1+f(d−1)]^W(1−f)^L, provided odds and the fraction remain constant and there are no other transactions.

Under these conditions, the percentage-staking ending balance depends on the numbers of wins and losses. Their order changes intermediate balances and drawdowns. Do not assume this order independence when odds vary, rules are reset, or constraints apply.

What happens during winning and losing streaks?

Percentage staking reduces the cash stake after losses and increases it after wins; flat staking does neither. Neither effect establishes a universal advantage.

Using the same $1000 starting bankroll, decimal odds of 2, and initial $20 stake:

After ten losses, the fixed $20 stake is 2.5% of the remaining bankroll. After ten wins, it is approximately 1.6667%. The percentage stake remains 2% in both cases.

At decimal odds of 2, one win followed by one loss multiplies the percentage-staking bankroll by 1.02 × 0.98 = 0.9996. Equal numbers of these wins and losses therefore do not restore it exactly to the starting amount. This is a consequence of multiplication—not evidence of a true 50% win probability or any particular median.

Recalculating stakes enables compounding through reinvestment. Unlike interest on a fixed-rate deposit, however, betting has uncertain outcomes rather than a fixed rate of return.

Recovery also depends on the changed base. A 20% loss requires a 25% gain on the remaining money to recover; a 30% loss requires a 42.857% gain. This applies regardless of the staking plan. The number of bets needed to recover is unknown in advance.

Drawdown is measured from a previous bankroll peak. A loss relative to the starting bankroll uses the starting amount as its base. These are different measures.

Why is average bankroll growth different from a typical result?

The mean ending bankroll, median, and expected logarithmic growth answer different questions. A higher mean does not promise a better result on your particular outcome sequence.

Suppose the win probability really is known and constant at p = 0.55, with decimal odds d = 2. The expected return edge per unit staked is e = pd−1 = 0.10, or 10%. The no-fee break-even probability is 1/d, here 50%. The five-percentage-point probability gap is not the same as the 10% expected return edge.

For a $1000 starting bankroll, fixed $20 stakes, a 2% bankroll fraction, and 100 independent bets, idealized calculations give:

The mean percentage-staking bankroll follows E[B] = B0(1+fe)^N, where N is the number of bets. This assumes independent binary outcomes, constant parameters, and no fees.

Darrell Velegol and Kyle J. M. Bishop's paper examines a mathematical model of sequential proportional bets and bankroll distributions. It supports conditional theory—not claims about observed bettor performance at sportsbooks.

An excessive fraction can harm growth even with positive EV. For example, with p = 0.51, d = 2, and f = 0.10, the expected return edge is 2%, but the expected logarithmic change per bet is negative:

g = 0.51 ln(1.10) + 0.49 ln(0.90) = −0.003018460972.

The full Kelly fraction in this example is 2%, but that is not a recommendation to stake 2%. An unknown or incorrectly estimated probability changes the conclusions. See the Kelly criterion guide for the approach's limitations. Entering an assumed probability into a calculator does not establish an edge.

How can you reproduce the calculator comparison?

You can reproduce the checked scenario in the staking plan calculator by enabling comparison, matching the initial cash stakes, and running the simulations.

1. Open the page and click Compare plans.
2. Enter 1000 in Bankroll.
3. Enter 2.00 in Odds.
4. Enter 55 in Win probability. This is an assumed probability expressed as a percentage, not a measured assessment of your forecasting skill.
5. Enter 100 in Bets, replacing the initial 500.
6. Enter 2 in Base unit. For this comparison, it sets the fixed cash stake at $20—2% of the starting bankroll.
7. Enter 2 in Percent. This sets each percentage stake at 2% of the current bankroll. The default values of 2% and 3% produce different initial stakes, so match them for this comparison.
8. Click Run 2500 simulations. Results appear after you click the button, not automatically.

Local comparison form with Bankroll 1000, Odds 2, Win probability 55%, Bets 100, Base unit 2, and Percent 2

Genuine screenshot of the English interface captured locally on localhost; it does not verify the production version. The 55% win probability is an assumption used for the calculation.

The calculation uses 2500 sequences and a reproducible random-generator seed of 174000, determined by the probability, odds, and number of bets. This makes the scenario repeatable; it does not validate the entered probability.

What do the results mean?

The ending-balance results are multiples of the starting bankroll, not net profit amounts. The checked scenario produces the following results for the first two plans:

The interface labels Worst 5% and Best 5% identify the fifth- and 95th-percentile boundaries. For example, 0.88× with a $1000 starting bankroll means approximately $880 at the lower boundary—not an 88% probability of profit or the average of all results in the lower tail.

1.20× means approximately $1200 in total, not $1200 in net profit. Matching rounded medians do not establish exact equality: the percentage plan's simulation-engine value is 1.197, displayed as 1.20×. It differs from the arithmetic mean of $1221.16.

Median drawdown is the median of the maximum peak-to-trough declines within individual sequences, not the decline in the ending bankroll. The percentile range describes a modeled distribution of results, not a confidence interval for your actual betting skill.

First two result rows showing Median end of 1.20× for both plans and Worst 5% boundaries of 0.88× and 0.87×

This cropped local interface capture shows only Plan, Median end, and Worst 5% for the same inputs. Ending balances are multiples of the starting bankroll; Best 5% and drawdown appear in the table above but are not visible in the capture.

The simulated Bust rate of 0.0% means only that no run reached an exact zero balance in this finite model sample. It does not mean zero practical risk: the balance can fall below a minimum stake or a chosen stopping threshold.

The calculator limits a stake to the remaining bankroll. It does not account for fees, pushes or void bets, real-world odds availability, minimum stakes, correlated outcomes, or overlapping bets. Matching the model's inputs does not guarantee matching behavior in real markets.

How should you choose a rule and define resets?

Choose a rule based on the losses you can afford, the quality of your probability estimates, and your ability to maintain accurate records—not a promise of the fastest bankroll growth. There is no universally optimal percentage, time horizon, or number of bets that proves a method effective.

Fixed cash stakes are simpler to track, but their share of the bankroll increases after losses. Percentage stakes automatically decrease as the base falls and increase as it grows, but require recalculation after settlement. With either method, check whether you can follow the rule as balances change, money is withdrawn, and minimum stakes apply.

A hybrid approach keeps the cash stake fixed between resets defined in advance. As an illustration—not a recommendation—you could start with 2% of $1000, or $20, and keep that amount until a reset condition is met. At a reset using a $1200 base, the new stake would be $24; using an $800 base, it would be $16. This is not recalculation after every bet.

Choose a review period or balance-based condition in advance. There is no basis for treating monthly or annual resets as best for everyone. Do not increase stakes after losses to chase recovery.

For each bet, record the stake, accepted odds, settlement result, actual net profit or loss, and current bankroll base. Track deposits and withdrawals separately. Do not include expected payouts in your bankroll or count money in open bets twice.

The bankroll management guide explains how to set aside betting funds and establish limits. For the separate question of sizing an individual bet, see what percentage of your bankroll to bet.

What can the questionnaire tell you?

The questionnaire can help identify a preference for fixed, percentage, or hybrid staking. It cannot establish which method will be more profitable or safer for you.

It asks about self-reported confidence, time horizon, bankroll size, risk preference, and a desired annual goal. Confidence and the goal are your answers, not measured evidence of an edge. The annual goal is not used in the current selection rule.

The result is a bounded preference heuristic. Any suggested 1–2% stake range has not been established as universally safe or optimal. Short- or long-horizon labels do not prove one method superior; the result is not an income promise or an investment plan.

Before applying any rule, define your bankroll base, acceptable total open exposure, and stopping condition. A staking plan controls the size of a potential loss; it does not turn an unknown probability into an edge.